Prediction of Drug resistance for target genetic variants

Genetic variants in drug targets can induce drug resistance. The cost of testing a single mutant using in vitro or in vivo experiments is unaffordable for highly variable targets (e.g. virus proteins).  When experimental and clinical data is available, machine learning methods can be used to model drug resistance response, but the reliability of predictions is very poor for untrained scenarios. Our approach is to perform high-accuracy prediction of affinities between disease-targets and drugs based on molecular simulation. There are a number of techniques that we can combine for these calculations: Ensemble docking, PELE with docking, and molecular mechanics simulation with Poisson-Boltzmann.